Effect of photoscale, interpreter bias and land type on woody crown-cover estimates from aerial photography
Woody vegetation cover interpreted from aerial photography requires assessment against field data as the signature of woody vegetation cover may differ between photoscales, vegetation types and photo-interpreters. Measurements of aerial woody cover taken from aerial photography of four different photoscales were compared with a field dataset from Eucalyptus- and Acacia-dominated landscapes of semi-arid Queensland. Two interpreters employed a method that utilises a stereoscope and sample-point graticule for manual quantified measurements of aerial woody cover. Both interpreters generated highly significant models accounting for 77 and 78% of deviance. Photoscale appears to have a consistent effect whereby the signature of woody cover increases as the photoscale decreases from 1 : 25 000 to 1 : 80 000, although the magnitude of this effect was different between interpreters. The results suggest no substantial differences in the shape of models predicting crown cover between Acacia- and Eucalyptus-dominated land types, although the precision of the models was greater for the Acacia (90–91% of residual deviance) than for the Eucalyptus (50–56% of residual deviance) land type. The reduced accuracy in the Eucalyptus land type probably reflects the relatively diffuse crowns of the dominant trees. The models generated for this dataset are within the range of those from other calibration studies employing photography of a range of scales and methodologies. The effect of photoscale is verified between the available studies, but there may also be variations arising from methodological differences or image properties. The present study highlights the influence of photoscale and interpreter bias for assessing woody crown cover from aerial photography. Studies that employ aerial photography should carefully consider potential biases and cater for them by calibrating assessments with field measurements.